The gravitational mass carried by sound waves

Angelo Esposito,1, 2, 3 Rafael Krichevsky,1 and Alberto Nicolis1

1
Department of Physics, Center for Theoretical Physics,
Columbia University, 538W 120th Street, New York, NY, 10027, USA
2
INFN, Sezione di Roma, Piazzale A. Moro 2, I-00185 Rome, Italy
3
Theoretical Particle Physics Laboratory (LPTP),
Institute of Physics, EPFL, Lausanne, Switzerland
We show that the commonly accepted statement that sound waves do not transport mass is only
true at linear order. Using effective field theory techniques, we confirm the result found in [1] for
zero-temperature superfluids, and extend it to the case of solids and ordinary fluids. We show
that, in fact, sound waves do carry mass—in particular, gravitational mass. This implies that a
sound wave not only is affected by gravity but also generates a tiny gravitational field, an aspect
arXiv:1807.08771v2 [gr-qc] 2 Mar 2019

not appreciated thus far. Our findings are valid for non-relativistic media as well, and could have
intriguing experimental implications.

PACS numbers: 46.40.Cd, 47.10.-g, 47.35.Rs, 47.37.+q

Keywords: Phonons, Sound Waves, Superfluids, Fluids, Solids

Introduction — It is usually said that sound waves do
not transport mass. They carry momentum and energy,
but it is an accepted fact that the net mass transported
by a sound wave vanishes. Here, we question this “fact”.
A first indication that sound waves can carry a nonzero
net mass is contained in the results of [1]: there, using an
effective point-particle theory, it was shown that phonons
in zero-temperature superfluids have an effective coupling
to gravity, which depends solely on their energy and on
the superfluid’s equation of state. For ordinary equations
of state, this coupling corresponds to a negative effective
gravitational mass: in the presence of an external gravi-
tational field, such as that of Earth, a phonon’s trajectory
bends upwards.
Now, this effect is completely equivalent to standard
refraction: in the presence of gravity, the pressure of the
superfluid depends on depth, and so does the speed of
sound. As a result, in the geometric acoustics limit sound
waves do not propagate along straight lines. Because of
this, one might be tempted to dismiss any interpreta-
tion of this phenomenon in terms of “gravitational mass”.
However, since in the formalism of [1] the effect is due to
a coupling with gravity in the effective Lagrangian of the
phonon, the same coupling must affect the field equation
for gravity: the (tiny) effective gravitational mass of the
phonon generates a (tiny) gravitational field. The source
of this gravitational field travels with the phonon. This
point was not appreciated in [1].
Thus, in a very physical sense, the phonon carries (neg-
ative) mass. Moreover, this is not due to the usual equiv-
alence of mass and energy in relativity: the effect survives
in the non-relativistic limit. And, finally, it is not a quan-
tum effect, because the formalism of [1] applies unaltered
to classical waves.
In this paper we confirm this result by computing ex-
plicitly the mass carried by a classical sound wave packet.
As we will see, from the wave mechanics standpoint the
fact that such a mass is nonzero is a non-linear effect, and
that is why from a linearized analysis we usually infer
that sound waves do not carry mass. We also generalize
the result to sound waves in ordinary fluids and to lon-
gitudinal and transverse sound waves in solids. We find
that, in the non-relativistic limit, the mass carried by a
sound wave traveling in these media is its energy E times
a factor that only depends on the medium’s equation of
state:
d log cs E
M =− , (1)
d log ρm c2s
where ρm is the medium’s mass density and cs the speed
of the sound wave under consideration. (For solids, the
correct formula is a minor generalization of this.) The
derivatives in Eq. (1) are standard adiabatic ones. Notice
that the same logarithmic derivative as in (1) appears in
the general formula for a sound wave’s radiation pressure
[2].
A final qualification: for excitations propagating in a
Poincaré-invariant vacuum, the invariant mass is a funda-
mental, completely unambiguous quantity, whose value
also directly determines the gap. Not so in a medium
that breaks boost invariance: there, the dispersion re-
lation E(p) of an excitation is not constrained by the
symmetries anymore, and, as a consequence, there is no
“invariant mass”. What do we mean by ‘M ’ in (1) then?
A non-relativistic medium—i.e. one with mass density
(times c2 ) much bigger than its energy density—has a
well-defined and conserved total mass. In this case, the
mass M is simply the fraction of this total mass that
travels with the excitation. It is perfectly consistent with
the symmetries of the system to have gapless excitations
(E(p) → 0 for p → 0) that nevertheless carry mass1 .
1 Because of this, we find the use of “massive” instead of “gapped”
in [3] somewhat inaccurate.
 2

For technical convenience, we use gravity to probe µ(t + π(x)). The e.o.m. associated with the action (2)
how much mass our excitations carry, since in the non- is ∂µ (P ′ (X)∂ µ φ) = 0, which expanded to quadratic or-
relativistic limit gravity only couples to mass. der in small fluctuations reads
Conventions: We set ~ = 1 but keep powers of c c2
 
(speed of light) explicit. We adopt a signature ηµν = π̈ − c2s ∇2 π = c2 1 − s2 π̇∇2 π
c
diag(−, +, +, +). Greek indices run over all spacetime (3)
c2s
   

2
coordinates, xµ = (ct, ~x), while Latin indices over spatial 2
+ ∂t c 1 − 2 ~ 2
∇π − g π̇ .
coordinates only. c

General framework and strategy — As a general The speed of sound cs and non-linear coupling g are given
framework, we use the recently developed effective field by combinations of derivatives of P (X), evaluated on the
theories (EFT) for the gapless excitations of generic me- background field φ(x) = µt. They can be rewritten in
dia. These can be characterized in terms of spontaneous terms of the chemical potential as c2s = c2 µ ddP/dµ
2 P/dµ2 and
symmetry breaking. In particular, all media break at c2

c2s

µ dcs
least part of the fundamental spacetime symmetries of g = 2c2 1 − c2 − cs dµ .
s

Nature, which are spacetime translations, spatial rota- We can solve this equation to non-linear order by writ-
tions, and Lorentz boosts (the Poincaré group) [4]. The ing π = π(1) + π(2) + . . . , where π(1) is a localized wave
Goldstone modes associated with the spontaneous break- packet obeying the linear equation of motion,
ing of these symmetries are gapless, and correspond to
the collective excitations of the system (see e.g. [5–20]). π̈(1) − c2s ∇2 π(1) = 0 , (4)
Here we compute the mass carried by a sound wave
and π(2) a quadratic correction, sourced by π(1) itself,
packet. For each of the media considered, we use the
corresponding effective action for gapless excitations to  
c2

π̈(2) − c2s ∇2 π(2) = ∂t c2 1 − 2s ~ (1) 2 − g π̇ 2


derive the equations of motion by varying the action ∇π (1)
c
w.r.t. the fields, and the energy-momentum tensor Tµν
2 2
  
by varying the action w.r.t. to the spacetime metric. c c
+ 2 1 − 2s π̇(1) 2
. (5)
Gravity interacts with any system through the system’s 2cs c
energy-momentum tensor, and, in particular, in the non-
To find the energy-momentum tensor, we vary the
relativistic limit the mass of a wave packet will be the
effective action (2) w.r.t. the metric, which enters the
spatial integral of c12 T00 evaluated on a solution of the
spacetime volume element and the invariant X =
equations of motion. Working perturbatively in the am-
−c2 g µν ∂µ φ∂ν φ 2 . The result is
plitude of our waves, we keep the linear terms and the
leading non-linear ones, since the effect we are interested Tµν = P (X)ηµν + 2c2 P ′ (X)∂µ φ∂ν φ . (6)
in vanishes at linear order.
Finally, we keep a relativistic notation and take the Expanding T00 to quadratic order in small fluctuations,
non-relativistic limit only when needed. It should be we find
noted that the same results can also be derived impos-
µnc2 c2
  
ing Galilean invariance from the beginning. However, the 2µcs dcs
T00 = 2 π̇ + 2 1 − 2 π̇ 2
relativistic analysis is not any more demanding to carry cs 2cs c dµ
(7)
out, and it is more general. Once powers of c are made c2

2c2
 
1 − 2s ~ 2 ,


explicit, the ratio of the speed of sound to c encodes how − ∇π
2 c
relativistic the system is. We will see that our effect sur-
vives when this ratio tends to zero, so it is not associated where n = dP/dµ is the number density.
with the usual relativistic equivalence of mass and energy. To find the mass transported by the sound wave, we
Zero-temperature superfluids — The low-energy dy- integrate T00 over a volume containing the wave packet
namics of a zero-temperature superfluid can be character- and average it over a time interval τ much larger than the
ized by a single scalar field φ(x), with low-energy effective typical oscillation time, τ ≫ 1/ω. The linear term in T00
action [9, 10] can receive contributions both from the linearized wave
Z packet π(1) and from the non-linear correction π(2) . The
S = dtd3 x P (X) , X ≡ −c2 ∂µ φ∂ µ φ , (2)

expanded about the ground state hφ(x)i = µt, where µ
is the (relativistic) chemical potential. The function P is
the pressure of the superfluid.
The superfluid phonon field π(x) corresponds to fluc-
tuations of the scalar around its background, φ(x) =
2 As usual, non-minimal couplings with the spacetime curvature
yield contributions to Tµν in flat space that are suppressed by
extra derivatives. In fact, they do not contribute at all to the
global charges—spatial integrals of T0µ —that we are trying to
compute.
 3

contribution from π(1) vanishes, since it is a total time solids has been developed in [13, 14, 19], and we will in
derivative and we are integrating on a time interval over large part follow their notation.
which the wave packet averages to zero. The contribution The EFT for a three-dimensional solid can be built out
from π(2) vanishes as well: the non-linear source on the of three scalar fields, φI (I = 1, 2, 3), representing the
r.h.s. of (5), in general, contains both oscillatory terms co-moving coordinates of the solid’s volume elements. In
as well as constant ones, but the time derivative gets rid the ground state, they can be aligned with the physical
of the latter. This means that π(2) is purely oscillatory as 1/3
coordinates, hφI (x)i = b0 xI , where b0 is a dimension-
well, and hence its time derivative vanishes in our time less free parameter representing the degree of compres-
average. sion/dilation of the solid. (The cubic root convention has
As far as the quadratic terms in T00 are concerned, only been chosen for later convenience.)
the linearized wave packet π(1) can contribute to this or-
The low energy effective action is a generic function of
der in perturbation theory. We can relate their volume
the combination B IJ ≡ ∂µ φI ∂ µ φJ that is invariant under
integral and time average to the energy E of our wave
the rotational symmetries of the solid, which act on the
packet, that is, the value of the free Hamiltonian. The
I, J indices. For simplicity, we focus on solids that are
Hamiltonian density that characterizes the ground state
isotropic at large distances3 . In such a case, there are
of a zero-temperature superfluid as well as its excitations
only three independent invariants that we can form out
is T 00 − µJ 0 , where J µ = 2P ′ (X)∂ µ φ is the Noether cur-
of B IJ . We find it convenient to work with
rent associated with particle-number conservation. We
thus find √ trB 2 trB 3
b≡ det B , Y ≡ 2 , Z≡ 3 . (12)
µnc2 1h
Z

i
~ 2 . (trB) (trB)
E= 2 d3 x π̇ 2 + c2s ∇π (8)
cs 2
The normalizations of Y and Z have been chosen so that
Representing
R the volume integral and time average with b is the only quantity that retains information about b0 —
h . . . i, and using the virial theorem, we have Y and Z are invariant under an overall rescaling of our
solid. The most general action for our solid is then
~ 2i = c2s
h π̇ 2 i = c2s h ∇π
R R

E. (9)
µnc2 Z
S = −w0 dtd3 x f (b, Y, Z) , (13)
Putting everything together,
R  µ dcs   ρ + P dcs  where w0 is a constant with the units of a mass density,
h T00 i = 1 − E = 1− E, (10)
cs dµ cs dP which we pulled out for convenience. It is convenient
to set w0 to be 1/c2 times the value of the relativistic
where in the last step we used the zero-temperature ther- enthalpy density ρ + P on the ground state. With such
modynamic identities ρ + P = µn, dP = ndµ. a normalization, on the background one has ∂f /∂b =
The above result is fully relativistic. In the non- c2 [14], which we implicity use below.
relativistic limit, the mass carried by our wave packet
A perfect fluid can be viewed as an infinitely symmet-
is
ric solid [4, 19], whose action is invariant under volume
1 R ρm dcs ρm dcs E preserving diffeomorphisms acting on the φI fields. Then,
M= h T00 i ≃ − E=− , (11)
c 2 cs dP cs dρm c2s the only invariant in this case is det B, and we can re-
cover the fluid EFT from the solid one by simply setting
which is indeed Eq. (1), as obtained in [1] from the point- to zero all derivatives of f with respect to Y or Z.
particle effective theory. It is simple to show that the Sound waves once again correspond to the fluctua-
same result can also be recovered in quantum field theory tions of the fields about their backgrounds, φI (x) =
for single-phonon states.
b0 1/3 xI + π I (x) . Expanding the invariants for small


Solids and fluids — As mentioned in the Introduction, fluctuations, one obtains the phonons’ quadratic equa-
the bending of a sound waves’s trajectory in the pres- tions of motion
ence of gravity is equivalent to Snell’s law. In light of
this, given the generality of Snell’s law, we expect sound π̈i − c2L ∂i ∇ ~ · ~π = ∂j Aij + ∂t Bi ,
~ · ~π − c2 ∇2 πi − ∂i ∇


T
waves to interact with gravity in fluids and solids much (14)
in the same way as they do in superfluids. We will again
analyze phonons in a solid from an effective field the-
ory viewpoint. Moreover, since in the EFT language a
fluid is simply a solid with an enhanced symmetry (see
e.g. [4, 19]), the results of this section extend straight- 3 Generalizing our analysis to a solid with discrete rotational sym-
forwardly to fluids as well. The effective field theory of metry is straightforward, although quite tedious.
 4

where The mass of a sound wave is therefore

2
Aij = g1 δij ~π˙ 2 + g2 π̇i π̇j + 3g3 δij ∇
~ · ~π R
M = w0 b0 h ∇~ · ~π i . (18)
+ g4 δij (∂k πℓ )2 + 2g4 ∂j πi ∇ ~ · ~π

2 The volume integral reduces to a boundary term and,
+ g5 δij ∇ ~ × ~π + 2g5 ∂j πi ∇ ~ · ~π (15)
for large enough volumes, one might be tempted to dis-
− 2g5 ∂i πj ∇ ~ · ~π + g6 ∂i πk ∂j πk + g6 ∂k πi ∂k πj card it. This is certainly allowed for a linearized solu-
+ g6 ∂k πi ∂j πk + 3g7 ∂i πk ∂k πj , tion, since we can take it to be as localized as we wish.
However, in general we must be careful about non-linear
and the vector Bi is a quadratic combination of fields corrections. Splitting as before the fluctuation field into
whose particular form is irrelevant for what follows4 . The ~π = ~π(1) + ~π(2) + . . . , from (14) one sees that for a local-
sound speeds and non-linear couplings are given by ized wave packet π(1) , the non-linear correction π(2) can
scale as 1/r2 at large distances, thus potentially giving a
16 fY + fZ finite contribution to (18).
c2L = b0 F ′′ + , (16a)
27 b0 To compute such a contribution, we first take the di-
4 fY + fZ vergence of (14), obtaining a wave equation directly for
c2T = , (16b)
9 b0 Ψ≡∇ ~ · ~π(2) , with ~π(1) -dependent sources:
1 c2 c2
g1 = − − L2 + T2 , (16c)
2 2c c ~ · B)
(∂t2 − c2L ∇2 )Ψ = ∂i ∂j Aij + ∂t (∇ ~ . (19)
c2
g2 = 1 − T2 , (16d)
c Time-averaging this equation, we get rid of the total time
c2 − c2T 1 10 16 fZ derivative terms:
g3 = L + b0 cL c′L − b0 cT c′T + , (16e)
3 3 9 243 b0 1
c2 8 fZ ∇2 hΨi = − h∂i ∂j Aij i , (20)
g4 = − L + 2b0 cT c′T − , (16f) c2L
2 27 b0
c2 − c2T 4 fZ where Aij is understood to be a function of the linear
g5 = L − b0 cT c′T + , (16g) solution ~π(1) . The resulting Poisson’s equation is solved
2 27 b0
2 fZ using Green’s function methods, ultimately leading to
g6 = c2T + , (16h) the volume-integrated and time-averaged expression we
9 b0
are after5 :
2 fZ
g7 = , (16i)
27 b0 ~ · ~π i ≃ − 1 h Aii i .
R R
h ∇ 2 (21)
3cL
where the prime, subscript Y , and subscript Z denote,
respectively, differentiation with respect to b, Y , and Z,
The right-hand side of (21) can be evaluated straight-
and all derivatives are evaluated on the background. No-
forwardly from the definitions in Eqs. (15) and (16). It is
tice that, with the exception of f ′ , the derivatives of f
convenient to split ~π(1) into longitudinal (L) and trans-
with respect to its arguments are of order c2L or c2T , and
verse (T ) components, for which the energies read
not of order c2 [14]. This is what makes the T00 below so
simple in the non-relativistic limit. Z
1 h˙2
i
Let us now find the gravitational mass carried by sound EL = w0 b0 d3 x ~ · ~πL 2 ,
~πL + c2L ∇ (22a)
2
waves in a solid (or fluid). It is convenient to take the
1 ˙2
Z h i
non-relativistic limit right away. The only O(c2 ) term in ET = w0 b0 d3 x ~ × ~πT 2 ,
~πT + c2T ∇


(22b)
T00 turns out to be 2

~ · ~π .
T00 ≃ c2 w0 b0 ∇ (17) and for which the obvious generalizations of the virial
relation (9) hold.
In particular, with our parametrization of fluctuations,
quadratic and higher-order terms contribute to the en-
ergy density of the solid, but not to its mass density.
5 To get to this, we used the distributional identity
1 1 4π 3
∂i ∂j = 3 (3r̂ i r̂ j − δij ) − x)δij ,
δ (~
4 r r 3
Note that we stopped differentiating between lower-case and
upper-case spatial indices, because the background configura- and the fact that the first piece—which should be interpreted
tions φI ∝ xI break spatial and internal rotations down to diago- as a 3D principal part, that is, by excising an infinitesimal ball
nal combinations, and so the equations of motion for fluctuations around r = 0—integrates to zero in any spherical region centered
will only be invariant under such combinations. at the origin.

Putting together Eqs. (18), (21) and (22) and using
the analogs of (9), in the non-relativistic limit we find
w0 b20
 

2
2
~ ~
′
R ′
R
M ≃ − 2 cL cL h ∇ · ~π i + cT cT h ∇ × ~π i
cL
 
b0 c′L c′T
=− 2 EL + ET . (23)
cL cL cT
Longitudinal and transverse sound waves in general
have different propagation speeds—in fact, one generally
has c2L > 43 c2T [21]. So even if initially one has a wave
packet made up of both components, over time these
will split into two spatially separated wave packets, each
propagating at its own speed. For each of these, one
simply has
d log cs E
M =− , (24)
d log ρm c2L
where cs is the propagation speed for the type of sound
wave in question (L or T ). We used the fact that in the
non-relativistic limit the mass density is proportional to
the level of compression, ρm ∝ b [22], and the derivative
with respect to log ρm has to be taken in the absence of
pure shear—that is, it corresponds to hydrostatic com-
pression [21]. (Using the standard relationships between
the stress and strain tensors in an isotropic solid, one
can also trade d/d log ρm for Kd/dP , where K is the
bulk modulus.) This generalizes Eq. (1) to the case of
both longitudinal and transverse waves.
This result holds equally well for the sound waves in a
perfect fluid, for which fY = fZ = c2T = 0, in which case
it reduces directly to Eq. (1). In fact, for a perfect fluid
we also checked our result by directly taking the non-
relativistic continuity and Euler equations for ρm and
~
a potential velocity field ~v = ∇ψ, expanding them to
quadratic order in the perturbations δρm and ψ about
a static
R and homogeneous background, and computing
M ≡ d3 x δρm for a perturbative solution of such equa-
tions of motion. We again obtain Eq. (1). We spare the
reader the tedious details of that analysis.
Conclusions — We showed that contrary to common
belief, sound waves carry gravitational mass in a standard
Newtonian sense: they are affected by gravity, but they
also source gravity. Our results show that this effect goes
hand-in-hand with the non-linear interactions of sound,
and that this is true for superfluids, fluids, and solids. In
particular, for all these media, in the non-relativistic limit
the mass transported by a sound wave is proportional to
its energy times a coefficient that only depends on the
medium’s equation of state.
The mass transported is in general quite small, of or-
der M ∼ E/c2s . For example, a very energetic phonon
in superfluid helium-4 with momentum k ∼ 1 keV (i.e.,
wavelength of order of the Bohr radius) carries a mass
roughly of order M ∼ 1 GeV, i.e. that of a single helium
5
atom. Nevertheless, it is possible to envision experimen-
tal setups where this effect could be detected.
One possibility is to employ ultra-cold atomic or molec-
ular gases. In these systems, not only might one be able
to achieve very small sound speeds and enhance the ef-
fect, but one could also use suitable traps to simulate
strong gravitational potentials [23]. For example, for
a cesium Bose-Einstein condensate, the typical sound
p
speed is cs ≃ 4πas n/m2Cs . The typical scattering
length and density are as ∼ 100 Å and n ∼ 1013 cm−3 ,
leading to cs ∼ 104 µm/s [24]. Assuming the maximum
phonon momentum to be (conservatively) k ∼ 1/as , one
gets a displaced mass of M ∼ 103 GeV. For a condensate
of radius ∼ 50 µm, this is roughly 10−3 ÷ 10−4 of the
total mass [25]. Although quite small, this could offer
interesting perspectives for detection.
Another possibility might be to consider seismic phe-
nomena. The wave generated by an earthquake of Richter
magnitude m = 9 carries an energy E ∼ 1018 Joules
which, for cs ∼ 5 km/s, corresponds to M ∼ 1011 kg, and
a change in gravitational acceleration δg ∼ 10−4 nm/s2 .
Atomic clocks and quantum gravimeters can currently
detect tiny changes in the gravitational acceleration, up
to fractions of nm/s2 [26–28]. Given the rapid improve-
ment of these techniques, one can imagine that in the not-
too-distant future they will reach the sensitivity needed
to detect the gravitational fields of seismic waves.
The effects we are considering may also be of rele-
vance to neutron star dynamics, since gravity would af-
fect phonon-mediated transport properties in the super-
fluid stellar interior [29, 30]. The sound speed inside the
core is expected to be relativistic [31, 32], but our results
hold nonetheless, with minor generalizations.
Finally, since sound waves both source gravity and
are affected by it, they can interact through gravity. In
particular, for ordinary equations of state (higher sound
speeds at higher pressures), their gravitational mass is
negative. Still, since gravity is attractive for like charges,
two sound wave packets running parallel to each other
should start converging. Very, very slowly, of course.
Acknowledgments — We are grateful to N. Bigagli,
G. Chiriacò, S. Garcia-Saenz, L. Hui, A. Lam, R. Mc-
nally, R. Penco, P. Richards, I. Z. Rothstein, C. Warner
and S. Will for illuminating discussions. This work has
been partially supported by the US Department of En-
ergy grant de-sc0011941. The work done by A.E. is also
supported by the Swiss National Science Foundation un-
der contract 200020-169696 and through the National
Center of Competence in Research SwissMAP.
[1] A. Nicolis and R. Penco, Phys. Rev. B97, 134516 (2018),
arXiv:1705.08914 [hep-th].
 6

[2] L. Brillouin, Ann. Phys. 10, 528 (1925). arXiv:1406.4422 [hep-th].

[3] H. Watanabe, T. Brauner, and H. Mu- [19] S. Endlich, A. Nicolis, and J. Wang, JCAP 1310, 49
rayama, Phys. Rev. Lett. 111, 021601 (2013), (2013).
arXiv:1303.1527 [hep-th]. [20] L. Alberte, M. Baggioli, V. C. Castillo, and O. Pujolas,
[4] A. Nicolis, R. Penco, F. Piazza, and R. Rattazzi, (2018), arXiv:1807.07474 [hep-th].
JHEP 06, 155 (2015), arXiv:1501.03845 [hep-th]. [21] L. Landau, E. Lifshitz, A. Kosevich, J. Sykes, W. Reid,
[5] R. V. Lange, PRL 14 (1965). and L. Pitaevskiı̆, Theory of Elasticity, Course of theo-
[6] H. Leutwyler, Helv. Phys. Acta 70, 275 (1997), retical physics (Butterworth-Heinemann, 1986).
arXiv:hep-ph/9609466 [hep-ph]. [22] S. Dubovsky, L. Hui, A. Nicolis, and
[7] R. V. Lange, Phys. Rev. 146 (1966). D. T. Son, Phys. Rev. D85, 085029 (2012),
[8] A. Nicolis and F. Piazza, arXiv:1107.0731 [hep-th].
Phys. Rev. Lett. 110, 011602 (2013), [Addendum: Phys. [23] S. Will, N. Bigagli, A. Lam, and C. Warner, private
Rev. Lett.110,039901(2013)], arXiv:1204.1570 [hep-th]. communication.
[9] F. Greiter, F. Wilczek, and E. Witten, Mod. Phys. Lett. [24] S. Will, private communication.
B, 903 (1989). [25] T. Weber, J. Herbig, M. Mark, H.-C. Nägerl, and
[10] D. T. Son, (2002), arXiv:hep-ph/0204199. R. Grimm, Science 299, 232 (2003).
[11] C. Hoyos, S. Moroz, and D. T. [26] C. Freier, M. Hauth, V. Schkolnik, B. Leykauf,
Son, Phys. Rev. B89, 174507 (2014), M. Schilling, H. Wziontek, H.-G. Scherneck, J. Müller,
arXiv:1305.3925 [cond-mat.quant-gas]. and A. Peters, in Journal of Physics: Conference Series,
[12] S. Moroz and C. Hoyos, Phys. Rev. B91, 064508 (2015), Vol. 723 (IOP Publishing, 2016) p. 012050.
arXiv:1408.5911 [cond-mat.quant-gas]. [27] C.-W. Chou, D. Hume, T. Rosenband, and D. Wineland,
[13] S. Dubovsky, T. Gregoire, A. Nicolis, and R. Rattazzi, Science 329, 1630 (2010).
JHEP 03, 025 (2006), arXiv:hep-th/0512260 [hep-th]. [28] S. L. Campbell, R. B. Hutson, G. E. Marti, A. Goban,
[14] S. Endlich, A. Nicolis, R. Rattazzi, and J. Wang, N. Darkwah Oppong, R. L. McNally, L. Sonderhouse,
JHEP 04, 102 (2011), arXiv:1011.6396 [hep-th]. J. M. Robinson, W. Zhang, B. J. Bloom, and J. Ye,
[15] M. Geracie and D. T. Son, JHEP 11, 004 (2014), Science 358, 90 (2017).
arXiv:1402.1146 [hep-th]. [29] A. B. Migdal, Sov. Phys. JETP 10, 176 (1960).
[16] G. Ballesteros, D. Comelli, and [30] V. L. Ginzburg and D. A. Kirzhnits, Sov. Phys. JETP
L. Pilo, Phys. Rev. D94, 025034 (2016), 20, 1346 (1965).
arXiv:1605.05304 [hep-th]. [31] Y. B. Zeldovich, Sov. Phys. JETP 14, 1143 (1961).
[17] G. Ballesteros, JCAP 1503, 001 (2015), [32] P. Haensel, A. Y. Potekhin, and D. G. Yakovlev, Neu-
arXiv:1410.2793 [hep-th]. tron Stars 1: Equation of State and Structure (Springer,
[18] B. Gripaios and D. Sutherland, 2007).
Phys. Rev. Lett. 114, 071601 (2015),